TABLE 2.

Linear mixed-effect modeling for multivariable affecting SARA, INAS, and SCAFI progression.

Response variable	Variables in fixed effect	Estimate	SE	95% CI	t	P
SARA	Time	1.491	0.079	1.336∼1.646	18.938	4.398e-50
	SARA at baseline	0.999	0.004	0.991∼1.007	245.779	0.000
	Time*Length of expanded allele	0.107	0.024	0.060∼0.154	4.493	1.1e-5
INAS	Time	0.547	0.046	0.457∼0.638	11.898	4.951e-26
	INAS at baseline	0.953	0.023	0.909∼0.998	42.149	1.298e-122
	Time*INAS at baseline	−0.099	0.020	−0.138∼–0.060	−4.999	1e-6
SCAFI	Time	−0.272	0.011	−0.293∼–0.251	−25.391	4.061e-60
	SCAFI at baseline	0.996	0.007	0.982∼1.010	139.931	3.340e-195
	Time*Length of normal allele	0.004	0.002	0.0002∼0.0071	2.060	0.041
	Time*SCAFI at baseline	−0.094	0.009	−0.111∼–0.076	−10.669	9.163e-21

Estimates derived from the model are given as mean, standard error (SE), and 95% confidence interval (CI); SARA, scale for the Assessment and Rating of Ataxia; SCAFI, SCA Functional Index; INAS, inventory of non-ataxia signs; SE, standard error. Length of expanded allele or normal allele refers to the CAG repeats in the ATXN3 gene. These three linear mixed models were built for exploring the independent factors which affects the SARA, INAS, or SCAFI progression. First, the univariate analysis was done separately for the seven interesting variables: repeat length of expanded allele, repeat length of normal allele, gender, age at baseline, age at onset, disease duration at baseline, and baseline scores of SARA or INAS or SCAFI. Independent factors that were significant in the univariate analysis were included and further tested in a multivariate model. These models also included family as a random effect. The detailed steps and data in constructing these three final models were shown in Supplementary Tables 2–4.